<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Sam Ansari on Medium]]></title>
        <description><![CDATA[Stories by Sam Ansari on Medium]]></description>
        <link>https://medium.com/@ansarisam?source=rss-9663a477bb27------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*i2rMONTMbKRGdc5qWnxfqw.jpeg</url>
            <title>Stories by Sam Ansari on Medium</title>
            <link>https://medium.com/@ansarisam?source=rss-9663a477bb27------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 02:55:10 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ansarisam/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The Hidden Cost of Contract Review — And How AI Is Changing Legal Operations]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://ansarisam.medium.com/the-hidden-cost-of-contract-review-and-how-ai-is-changing-legal-operations-dadcb4412fc0?source=rss-9663a477bb27------2"><img src="https://cdn-images-1.medium.com/max/1254/1*5dm2crFBM10TGOdocK3Mog.png" width="1254"></a></p><p class="medium-feed-snippet">Most organizations treat contracts as legal documents.</p><p class="medium-feed-link"><a href="https://ansarisam.medium.com/the-hidden-cost-of-contract-review-and-how-ai-is-changing-legal-operations-dadcb4412fc0?source=rss-9663a477bb27------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://ansarisam.medium.com/the-hidden-cost-of-contract-review-and-how-ai-is-changing-legal-operations-dadcb4412fc0?source=rss-9663a477bb27------2</link>
            <guid isPermaLink="false">https://medium.com/p/dadcb4412fc0</guid>
            <category><![CDATA[legaltech]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-contract-review]]></category>
            <dc:creator><![CDATA[Sam Ansari]]></dc:creator>
            <pubDate>Tue, 23 Jun 2026 13:33:45 GMT</pubDate>
            <atom:updated>2026-06-23T13:33:45.685Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Using Generative AI in Regulatory Compliance Auditing]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://ansarisam.medium.com/using-generative-ai-in-regulatory-compliance-auditing-32ab3876bbd8?source=rss-9663a477bb27------2"><img src="https://cdn-images-1.medium.com/max/2600/1*kdwt0tbDddOZMCpD4CtLaA.png" width="2784"></a></p><p class="medium-feed-snippet">Regulatory compliance auditing has traditionally been a manual, time-intensive process involving the interpretation of complex regulations&#x2026;</p><p class="medium-feed-link"><a href="https://ansarisam.medium.com/using-generative-ai-in-regulatory-compliance-auditing-32ab3876bbd8?source=rss-9663a477bb27------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://ansarisam.medium.com/using-generative-ai-in-regulatory-compliance-auditing-32ab3876bbd8?source=rss-9663a477bb27------2</link>
            <guid isPermaLink="false">https://medium.com/p/32ab3876bbd8</guid>
            <category><![CDATA[ai-powered-compliance]]></category>
            <category><![CDATA[modernizing-compliance-ai]]></category>
            <category><![CDATA[generative-ai-for-audit]]></category>
            <category><![CDATA[smarter-compliance]]></category>
            <category><![CDATA[ai-in-compliance-audits]]></category>
            <dc:creator><![CDATA[Sam Ansari]]></dc:creator>
            <pubDate>Thu, 29 May 2025 04:19:45 GMT</pubDate>
            <atom:updated>2025-05-29T04:19:45.359Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Augmenting Compliance Teams with AI: Efficiency, Scale, and Measurable ROI]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://ansarisam.medium.com/augmenting-compliance-teams-with-ai-efficiency-scale-and-measurable-roi-9fdfea51b052?source=rss-9663a477bb27------2"><img src="https://cdn-images-1.medium.com/max/2042/1*sFTK_CmCpMc-GeuBAwR8_w.png" width="2042"></a></p><p class="medium-feed-snippet">Scale your compliance practice and unlock up to $1.8M in ROI&#x200A;&#x2014;&#x200A;without increasing headcount.</p><p class="medium-feed-link"><a href="https://ansarisam.medium.com/augmenting-compliance-teams-with-ai-efficiency-scale-and-measurable-roi-9fdfea51b052?source=rss-9663a477bb27------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://ansarisam.medium.com/augmenting-compliance-teams-with-ai-efficiency-scale-and-measurable-roi-9fdfea51b052?source=rss-9663a477bb27------2</link>
            <guid isPermaLink="false">https://medium.com/p/9fdfea51b052</guid>
            <category><![CDATA[compliance-automation]]></category>
            <category><![CDATA[compliance-management]]></category>
            <category><![CDATA[aiincompliance]]></category>
            <category><![CDATA[compliance]]></category>
            <category><![CDATA[ai-for-compliance]]></category>
            <dc:creator><![CDATA[Sam Ansari]]></dc:creator>
            <pubDate>Wed, 21 May 2025 01:47:36 GMT</pubDate>
            <atom:updated>2025-05-25T16:11:18.563Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[AI and Immortality: Science Fiction or Future Reality?]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-snippet">Introduction</p><p class="medium-feed-link"><a href="https://ansarisam.medium.com/ai-and-immortality-science-fiction-or-future-reality-61cb72d9d3a6?source=rss-9663a477bb27------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://ansarisam.medium.com/ai-and-immortality-science-fiction-or-future-reality-61cb72d9d3a6?source=rss-9663a477bb27------2</link>
            <guid isPermaLink="false">https://medium.com/p/61cb72d9d3a6</guid>
            <category><![CDATA[microship-and-ai]]></category>
            <category><![CDATA[ai-extending-life]]></category>
            <category><![CDATA[ai-for-life]]></category>
            <category><![CDATA[ai-and-medical]]></category>
            <category><![CDATA[human-immortability]]></category>
            <dc:creator><![CDATA[Sam Ansari]]></dc:creator>
            <pubDate>Fri, 21 Mar 2025 02:50:18 GMT</pubDate>
            <atom:updated>2025-03-21T02:50:18.429Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Ethical Considerations of AI: Balancing Innovation and Responsibility]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://ansarisam.medium.com/ethical-considerations-of-ai-balancing-innovation-and-responsibility-c3edd4dce35d?source=rss-9663a477bb27------2"><img src="https://cdn-images-1.medium.com/max/1024/1*7OPbzhIitnaHsbm65IMF2Q.jpeg" width="1024"></a></p><p class="medium-feed-snippet">Artificial Intelligence (AI) has become an integral part of modern society, influencing industries such as healthcare, finance, education&#x2026;</p><p class="medium-feed-link"><a href="https://ansarisam.medium.com/ethical-considerations-of-ai-balancing-innovation-and-responsibility-c3edd4dce35d?source=rss-9663a477bb27------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://ansarisam.medium.com/ethical-considerations-of-ai-balancing-innovation-and-responsibility-c3edd4dce35d?source=rss-9663a477bb27------2</link>
            <guid isPermaLink="false">https://medium.com/p/c3edd4dce35d</guid>
            <category><![CDATA[social-ai]]></category>
            <category><![CDATA[responsible-ai]]></category>
            <category><![CDATA[ai-and-humans]]></category>
            <category><![CDATA[ethical-ai]]></category>
            <dc:creator><![CDATA[Sam Ansari]]></dc:creator>
            <pubDate>Fri, 14 Mar 2025 01:25:45 GMT</pubDate>
            <atom:updated>2025-03-14T01:25:45.848Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[1,000 Industrial Use Cases of Artificial Intelligence]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://ansarisam.medium.com/1-000-industrial-use-cases-of-artificial-intelligence-7328013240c8?source=rss-9663a477bb27------2"><img src="https://cdn-images-1.medium.com/max/1000/0*IgvhmcLBgsm-h9CQ.jpg" width="1000"></a></p><p class="medium-feed-snippet">Introduction</p><p class="medium-feed-link"><a href="https://ansarisam.medium.com/1-000-industrial-use-cases-of-artificial-intelligence-7328013240c8?source=rss-9663a477bb27------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://ansarisam.medium.com/1-000-industrial-use-cases-of-artificial-intelligence-7328013240c8?source=rss-9663a477bb27------2</link>
            <guid isPermaLink="false">https://medium.com/p/7328013240c8</guid>
            <dc:creator><![CDATA[Sam Ansari]]></dc:creator>
            <pubDate>Thu, 06 Mar 2025 13:21:18 GMT</pubDate>
            <atom:updated>2025-03-06T13:21:18.368Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Machine Learning and AI Without Coding]]></title>
            <link>https://ansarisam.medium.com/machine-learning-and-ai-without-coding-3fb15c5e5d03?source=rss-9663a477bb27------2</link>
            <guid isPermaLink="false">https://medium.com/p/3fb15c5e5d03</guid>
            <dc:creator><![CDATA[Sam Ansari]]></dc:creator>
            <pubDate>Mon, 28 Oct 2024 02:36:56 GMT</pubDate>
            <atom:updated>2024-10-28T02:36:56.839Z</atom:updated>
            <content:encoded><![CDATA[<p>Machine learning, a subset of artificial intelligence (AI), has become a buzzword in today’s digital age, transforming industries and automating tasks. However, the perception that machine learning requires extensive programming knowledge can be a barrier for many. This book “AI for Non-Programmers: Building Data Science and AI Solutions Without Coding,” demystifies machine learning and empowers non-programmers to harness its power. The book is available on Amazon, in paperback, hardcover, and kindle formats, at <a href="https://a.co/d/igy4L1N">https://a.co/d/igy4L1N</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/365/1*KZW2PwwshAgLvDpF0PhY-Q.jpeg" /><figcaption><a href="https://a.co/d/a2Mt5GW">https://a.co/d/a2Mt5GW</a></figcaption></figure><h3>For Non-Programmers</h3><p>This book is designed for individuals who wish to understand and apply machine learning concepts without the need for intricate coding. It is an essential read for business analysts, marketers, managers, and anyone with a passion for data-driven insights. The author’s mission is to make machine learning accessible to everyone, regardless of their programming background.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/365/1*cRsL05vEcuVi77UW_sDdrg.jpeg" /><figcaption><a href="https://a.co/d/1TO4WD2">https://a.co/d/1TO4WD2</a></figcaption></figure><h3>Introduction to Machine Learning</h3><p>The book begins by introducing machine learning as a technology that enables computers to learn from data and make decisions or predictions without being explicitly programmed. It covers the four main types of machine learning: supervised learning, unsupervised learning, reinforcement learning, and deep learning. The book provides real-world examples and practical applications of each type, making the concepts easy to understand.</p><h3>No-code Platform: Momentum</h3><p>Another crucial aspect of the book is the utilization of no-code platform, Accure Momentum, <a href="https://accure.ai/momentum">https://accure.ai/momentum</a>. This user-friendly platform empower users to build machine learning models, automate processes, and extract insights from data with intuitive drag-and-drop interfaces. The book also explores these tools in depth, ensuring readers have the skills they need to succeed.</p><h3>Machine Learning Lifecycle</h3><p>The book also covers the entire machine learning development lifecycle, from data engineering and model training to deployment, monitoring, and management. It emphasizes the importance of data engineering, which plays a crucial role in preparing data for machine learning models. The book covers key concepts in data engineering, such as SQL, data types, and data sources.</p><h3>1000 Industrial AI Use Cases</h3><p>Furthermore, the book provides a comprehensive list of 1,000 AI use cases across 20 industries, including healthcare, finance, retail, manufacturing, and more. Each use case includes suitable algorithms, making it an invaluable resource for identifying practical applications of machine learning within various organizations or areas of interest.</p><h3>Practical Guide</h3><p>This book is not just a theoretical guide; it is a practical one. It emphasizes hands-on, real-world scenarios and projects, allowing readers to apply machine learning concepts without the need for intricate coding. The author covers various machine learning tasks, such as predictive analytics, recommendation systems, and data engineering.</p><h3>Conclusion</h3><p>“From Concept to Creation: Building Data Science and AI Solutions Without Coding: A Practical Guide for Non-Programmers, Subject Matter Experts, Engineers, Students, and Analysts” is an essential read for anyone looking to understand and apply machine learning concepts without the need for extensive programming knowledge. Shamshad Ansari’s book breaks down the barriers to entry, making machine learning accessible to business leaders, managers, and professionals, enabling them to become more data-savvy, make informed decisions, and drive innovation within their organizations.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3fb15c5e5d03" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How Generative AI Enables Semantic Graphs for Knowledge Extraction]]></title>
            <link>https://ansarisam.medium.com/how-generative-ai-enables-semantic-graphs-for-knowledge-extraction-77478eb58108?source=rss-9663a477bb27------2</link>
            <guid isPermaLink="false">https://medium.com/p/77478eb58108</guid>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[securegpt]]></category>
            <category><![CDATA[knowledge-extraction]]></category>
            <category><![CDATA[semantic-graph]]></category>
            <dc:creator><![CDATA[Sam Ansari]]></dc:creator>
            <pubDate>Mon, 09 Sep 2024 05:17:54 GMT</pubDate>
            <atom:updated>2025-03-03T01:54:05.663Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*khYKOaWhAP8xkmu_JVqzGw.png" /><figcaption>Semantic knowledge graph</figcaption></figure><p>Generative AI has rapidly evolved as a game-changing technology, capable of synthesizing and understanding vast amounts of data across different formats. One particularly powerful application of generative AI is in enabling the creation and utilization of semantic graphs for knowledge extraction. By transforming unstructured data into structured, interconnected knowledge representations, generative AI is advancing how we process and interpret information in a wide range of fields.</p><p>This article explores how generative AI empowers the development of semantic graphs and how these graphs, in turn, facilitate deeper knowledge extraction and understanding across various domains.</p><h3>What is Generative AI?</h3><p>Generative AI refers to artificial intelligence systems capable of generating content, whether text, images, or even data structures, based on learned patterns. Unlike traditional AI, which is more focused on classification or prediction, generative AI models can create new data that resembles the input they were trained on. Popular examples include models like GPT (Generative Pre-trained Transformer).</p><p>In the context of semantic graph creation, generative AI models can analyze unstructured data and generate structured representations of the entities and their relationships, effectively building semantic graphs that can be used for knowledge extraction.</p><h3>Understanding Semantic Graphs and Knowledge Extraction</h3><p>A semantic graph is a data structure that represents real-world entities (such as people, objects, or concepts) and the relationships between them. It is widely used in knowledge extraction to create a network of information where entities are nodes, and their relationships are edges. These graphs offer a powerful way to capture the meaning and context of information.</p><p>By visualizing data in this interconnected format, organizations can uncover complex relationships and gain deeper insights that would otherwise remain hidden in raw, unstructured data. This is where generative AI comes into play — by automating the extraction and mapping of information into a semantic graph format, it enables faster and more accurate knowledge extraction.</p><h3>How Generative AI Enables Semantic Graph Creation</h3><h3>1. Automatic Entity Recognition and Extraction</h3><p>Generative AI models can be trained to recognize and extract entities from unstructured data sources like text, video, or audio. These entities, such as people, organizations, products, or locations, form the nodes in a semantic graph. For example, a generative AI model could process a news article and automatically identify key individuals, events, and organizations involved in the story.</p><p>This automation reduces the time and effort needed to manually identify entities in large datasets, providing a significant advantage for businesses and researchers dealing with vast amounts of data.</p><h3>2. Generating Relationships Between Entities</h3><p>Once entities are identified, the next step is to understand how they are related. Generative AI can automatically infer and generate relationships between entities by analyzing patterns in the data. These relationships, such as “works for,” “is located in,” or “is part of,” form the edges of the semantic graph.</p><p>Generative models are particularly useful here because they can go beyond surface-level analysis and infer deeper, often implicit relationships. For example, in a corporate email dataset, a generative AI model might recognize not only direct collaborations between individuals but also infer informal networks of influence based on email tone and frequency.</p><h3>3. Contextual Understanding and Knowledge Mapping</h3><p>Context is critical for accurate knowledge extraction. Generative AI can process data in a way that captures the contextual meaning of entities and relationships, ensuring that the semantic graph reflects not just isolated facts but also the nuances of how those facts interrelate.</p><p>For instance, in healthcare, generative AI can be used to build semantic graphs from medical records and clinical data. The model can identify not only individual symptoms, treatments, and diagnoses but also how these elements are interconnected — perhaps revealing a correlation between a specific treatment and better patient outcomes in certain contexts.</p><h3>4. Cross-Domain Knowledge Integration</h3><p>One of the powerful advantages of generative AI is its ability to integrate knowledge across multiple domains. For example, a generative AI model can analyze both text and video data, combining insights from each into a unified semantic graph. In media analysis, this might involve extracting entities from both a news article and a related video interview, merging them to provide a more holistic understanding of a story.</p><p>This ability to pull in information from various sources and formats significantly enhances the breadth and depth of the knowledge represented in the semantic graph.</p><h3>Demonstrating Semantic Graph Generation from Video Using SecureGPT</h3><p>In today’s AI-driven world, the ability to extract knowledge from multimedia content like videos is becoming increasingly valuable. SecureGPT, a generative AI platform, offers a powerful solution for transforming video content into structured, insightful data through semantic graph generation. In this article, I demonstrate how SecureGPT processes video content to generate semantic graphs, turning complex, unstructured information into actionable insights.</p><h3>How Semantic Graphs are Generated from Video Using SecureGPT</h3><p>In this demonstration, SecureGPT takes a video file as input, containing both visual and audio components, and processes the data to generate semantic graphs. A semantic graph is a representation of entities (such as people, objects, or concepts) and the relationships between them. By using SecureGPT, users can extract meaningful relationships and contextual information from video data, providing valuable insights into the content.</p><p>Here is the link to the video: <a href="https://youtu.be/caLihAALHRk">https://youtu.be/caLihAALHRk</a></p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FcaLihAALHRk%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DcaLihAALHRk&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FcaLihAALHRk%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/1b6d70f779a1522a545c051aadd18374/href">https://medium.com/media/1b6d70f779a1522a545c051aadd18374/href</a></iframe><h3>How Semantic Graphs are Generated from Video Using SecureGPT</h3><p>In this demonstration, SecureGPT takes a video file as input, containing both visual and audio components, and processes the data to generate semantic graphs. A semantic graph is a representation of entities (such as people, objects, or concepts) and the relationships between them. By using SecureGPT, users can extract meaningful relationships and contextual information from video data, providing valuable insights into the content.</p><h3>SecureGPT Output: Customizable Knowledge Representation</h3><p>Once SecureGPT has analyzed the video, the output can be delivered in several formats based on user specifications, including triples, JSON, or any other structured data format. Here’s a breakdown of the available output options:</p><ol><li><strong>Triples</strong>: One common way to represent relationships in a semantic graph is through triples. Each triple consists of a subject, predicate, and object, essentially forming a statement about the relationships between entities. For example, a video might generate the following triples:</li></ol><pre>(Person1, interacts_with, Person2)<br>(Vehicle, located_at, StreetName)<br>(Speaker, mentions, ProductName)</pre><p><strong>2. JSON</strong>: SecureGPT can also output the semantic graph in JSON format, which is ideal for integration with other systems or for further analysis. The JSON structure might look like this:</p><pre>{<br>   &quot;entities&quot;: [<br>      {&quot;name&quot;: &quot;Person1&quot;, &quot;type&quot;: &quot;human&quot;},<br>      {&quot;name&quot;: &quot;Person2&quot;, &quot;type&quot;: &quot;human&quot;},<br>      {&quot;name&quot;: &quot;Vehicle&quot;, &quot;type&quot;: &quot;object&quot;}<br>   ],<br>   &quot;relationships&quot;: [<br>      {&quot;subject&quot;: &quot;Person1&quot;, &quot;predicate&quot;: &quot;interacts_with&quot;, &quot;object&quot;: &quot;Person2&quot;},<br>      {&quot;subject&quot;: &quot;Vehicle&quot;, &quot;predicate&quot;: &quot;located_at&quot;, &quot;object&quot;: &quot;StreetName&quot;}<br>   ]<br>}</pre><p><strong>3. Custom Formats</strong>: SecureGPT allows users to define custom output formats depending on their specific needs. This flexibility ensures that the output is easily interpretable and can be integrated into workflows seamlessly.</p><h3>Example: How SecureGPT Works in a Video Scenario</h3><p>Imagine you have a video of a business conference where different speakers present products and ideas. SecureGPT can process this video to generate a semantic graph that captures the following:</p><ul><li><strong>Entity Recognition</strong>: The platform identifies key entities such as the speakers, audience members, products mentioned, and locations within the conference.</li><li><strong>Relationship Extraction</strong>: SecureGPT generates relationships between these entities, such as which speaker discussed which product, or which attendee asked questions during the Q&amp;A session.</li><li><strong>Contextual Understanding</strong>: By processing both the video and audio, SecureGPT can also analyze the context in which certain products or ideas were presented, providing deeper insights into the conversation.</li></ul><p>The resulting output could be a set of triples or a JSON file detailing these interactions. For example, SecureGPT might generate a triple like (Speaker1, presents, ProductX) and further enrich the graph with relationships like (AudienceMember, asks_question_about, ProductX).</p><h3>Applications of Generative AI-Powered Semantic Graphs</h3><p>Generative AI-powered semantic graphs have the potential to transform a wide array of industries by enabling faster, more accurate knowledge extraction and analysis. Here are a few key examples:</p><h3>1. Healthcare</h3><p>Generative AI can extract knowledge from unstructured medical data, including patient records, research papers, and medical imaging. By building semantic graphs that represent diseases, treatments, and outcomes, healthcare providers can improve diagnosis accuracy, discover new treatment options, and better understand patient journeys.</p><p>For example, a semantic graph might reveal patterns between patient demographics, treatment plans, and recovery rates, offering valuable insights for precision medicine and personalized healthcare strategies.</p><h3>2. Legal and Compliance</h3><p>Legal documents are often dense and filled with complex language, making it difficult to extract relevant information. Generative AI can automate the extraction of legal entities (such as parties involved in a case, contracts, or regulatory statutes) and their relationships. This makes it easier for legal professionals to search, analyze, and draw insights from vast libraries of documents.</p><p>A semantic graph could highlight relationships between court cases, legal precedents, and specific statutes, helping lawyers quickly find relevant information.</p><h3>3. Financial Services</h3><p>In finance, understanding the relationships between market events, companies, and economic indicators is critical. Generative AI can build semantic graphs that link financial reports, market news, and macroeconomic trends, giving investors and analysts a clearer picture of market dynamics.</p><p>For instance, an AI-generated semantic graph could connect corporate earnings reports with stock price fluctuations and broader market movements, helping analysts identify investment opportunities or risks.</p><h3>4. Business Intelligence</h3><p>Generative AI can help companies build semantic graphs that represent internal data, including sales figures, customer interactions, and supply chain information. By analyzing these relationships, businesses can optimize operations, improve customer experiences, and make data-driven decisions more efficiently.</p><p>For example, a company might use a semantic graph to map the relationship between customer satisfaction scores and product delivery times, uncovering areas for operational improvements.</p><h3>Conclusion</h3><p>Generative AI is transforming how semantic graphs are created and utilized for knowledge extraction. By automating the identification of entities, generating meaningful relationships, and understanding the context in which information exists, generative AI enables faster and more insightful analysis of unstructured data. Whether in healthcare, finance, legal services, or other domains, the synergy between generative AI and semantic graphs is unlocking new possibilities for extracting and interpreting knowledge from diverse data sources.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=77478eb58108" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How MOMENTUM Enables LLMOps]]></title>
            <link>https://ansarisam.medium.com/how-momentum-enables-llmops-c6364aac7ccc?source=rss-9663a477bb27------2</link>
            <guid isPermaLink="false">https://medium.com/p/c6364aac7ccc</guid>
            <category><![CDATA[operationalizing-llm]]></category>
            <category><![CDATA[momentum-for-llm]]></category>
            <category><![CDATA[fine-tuning-llm]]></category>
            <category><![CDATA[data-for-llm-finetuning]]></category>
            <category><![CDATA[mlops-deployment]]></category>
            <dc:creator><![CDATA[Sam Ansari]]></dc:creator>
            <pubDate>Sat, 06 Jul 2024 22:35:52 GMT</pubDate>
            <atom:updated>2024-07-06T22:35:52.883Z</atom:updated>
            <content:encoded><![CDATA[<p>Large Language Model Operations (LLMOps) is a crucial aspect of deploying, managing, and optimizing large language models (LLMs) in real-world applications. Accure’s MOMENTUM platform simplifies and accelerates this process by providing a robust, no-code data engineering and AI/ML environment. This white paper explores how MOMENTUM facilitates LLMOps through data preparation, fine-tuning, testing, deployment, and interactive use, along with practical business use cases.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/903/1*kQwbL09ynDwO4cFlDQGwqw.png" /><figcaption>MOMENTUM Architecture</figcaption></figure><h3>Data Preparation with MOMENTUM</h3><p>Effective LLMOps begins with comprehensive data preparation. MOMENTUM streamlines this phase with its powerful data ingester, transformer, and data pipeline tools.</p><h3>Data Ingester</h3><p>MOMENTUM’s data ingester allows users to seamlessly import data from various sources, including databases, cloud storage, and APIs. The intuitive drag-and-drop interface ensures that even non-technical users can easily set up data ingestion workflows. This capability ensures that all necessary data is readily available for the subsequent stages of model development.</p><h3>Data Transformer</h3><p>Once the data is ingested, the transformer module provides a suite of tools for data cleaning, normalization, and augmentation. Users can apply transformations such as tokenization, stemming, and lemmatization, which are essential for preparing text data for LLMs. The transformer module also supports custom scripts, giving users the flexibility to tailor data preparation to their specific needs.</p><h3>Data Pipeline</h3><p>MOMENTUM’s data pipeline orchestrates the entire data flow, from ingestion to transformation. Users can define complex workflows that include branching, conditional logic, and scheduling. The platform’s visual interface makes it easy to monitor and manage data pipelines, ensuring data consistency and reliability throughout the model development lifecycle.</p><h3>LLM Fine-Tuning with MOMENTUM’s ML/NLP Module</h3><p>Fine-tuning is a critical step in adapting pre-trained LLMs to specific tasks. MOMENTUM’s ML/NLP module offers robust capabilities for fine-tuning models with minimal effort.</p><h3>Pre-trained Model Integration</h3><p>MOMENTUM provides seamless integration with popular pre-trained LLMs available on Huggingface, such as Llama, Mistral, and Falcon. Users can select a model from the library and fine-tune it on their custom dataset. The platform supports various fine-tuning techniques, including supervised learning, reinforcement learning, and transfer learning.</p><h3>Custom Training Workflows</h3><p>The ML/NLP module enables users to define custom training workflows using a visual interface. Users can specify hyperparameters, select optimization algorithms, and configure training schedules. MOMENTUM also supports distributed training, allowing users to leverage multiple GPUs and accelerators for faster fine-tuning.</p><h3>Testing and Evaluation with MOMENTUM’s ML Prediction Module</h3><p>After fine-tuning, rigorous testing and evaluation are essential to ensure model performance and reliability. MOMENTUM’s ML Prediction module provides comprehensive tools for this purpose.</p><h3>Model Testing</h3><p>The platform offers a range of testing techniques, including cross-validation, holdout validation, and k-fold validation. Users can evaluate their models on various metrics such as accuracy, precision, recall, F1 score, ROUGE score, and BLEU score. MOMENTUM also supports A/B testing, enabling users to compare the performance of different models or configurations.</p><h3>Model Evaluation</h3><p>MOMENTUM’s evaluation tools include confusion matrices, ROC curves, precision-recall curves, ROUGE scores, and BLEU scores, providing deep insights into model behavior. Users can analyze model performance on different subsets of data, helping identify potential biases or weaknesses. The platform also generates detailed evaluation reports, facilitating easy communication of results to stakeholders.</p><h3>Model Deployment and Inference with MOMENTUM’s MLOps</h3><p>Deploying LLMs for inference requires robust MLOps capabilities. MOMENTUM simplifies this process by automating key aspects of deployment and providing advanced monitoring and management features.</p><h3>Secure Inference Endpoint</h3><p>MOMENTUM automatically creates secure inference endpoint URLs, ensuring that deployed models are accessible only to authorized users. The platform supports role-based access control, allowing administrators to define and manage user permissions. This ensures that sensitive models and data are protected from unauthorized access.</p><h3>Feature Store</h3><p>The integrated feature store allows users to manage and reuse features across different models and projects. This reduces duplication of effort and ensures consistency in feature engineering. The feature store also supports versioning, making it easy to track and revert changes.</p><h3>Data Drift Detection and Monitoring</h3><p>MOMENTUM continuously monitors deployed models for data drift, ensuring that models remain accurate and relevant over time. The platform provides real-time alerts and dashboards, allowing users to quickly identify and address issues. MOMENTUM’s monitoring capabilities also include performance tracking, enabling users to assess model latency, throughput, and other critical metrics.</p><h3>Interactive User Interface with SecureGPT</h3><p>Finally, MOMENTUM offers a SecureGPT web interface, allowing users to interactively chat with and ask questions to deployed models. This interface is designed for ease of use, providing a conversational experience that enhances user engagement and accessibility.</p><h3>Secure and User-Friendly</h3><p>The SecureGPT interface is built with security in mind, ensuring that user interactions are encrypted and protected. The intuitive design allows users to ask questions, get answers, and provide feedback, facilitating continuous improvement of deployed models.</p><h3>Real-Time Interaction</h3><p>Users can engage with models in real-time, exploring their capabilities and performance. This interactive approach not only enhances user satisfaction but also provides valuable insights into model behavior, informing future improvements and updates.</p><h3>Business Use Cases</h3><h3>Customer Support Automation</h3><p>A company can use MOMENTUM to develop and deploy a customer support chatbot powered by an LLM fine-tuned on their specific product information and customer service guidelines. The chatbot can handle common queries, process returns, and provide product recommendations, significantly reducing the load on human agents and improving customer satisfaction.</p><h3>Personalized Marketing</h3><p>Marketing agencies can leverage MOMENTUM to fine-tune a pre-trained LLM for generating personalized email content. By integrating customer behavior and preference data, agencies can create targeted marketing campaigns that result in higher engagement and conversion rates. The deployment of these models using MOMENTUM’s MLOps features ensures secure and efficient delivery.</p><h3>Healthcare Document Analysis</h3><p>Healthcare providers can use MOMENTUM to fine-tune an LLM for analyzing and summarizing patient records and clinical trial documents. The model can help medical professionals quickly access relevant information, improving decision-making and patient outcomes. Continuous monitoring for data drift and performance metrics ensures the model remains accurate and reliable over time.</p><h3>Financial Fraud Detection</h3><p>Financial services firms can implement an LLM fine-tuned using MOMENTUM to detect fraudulent activities in transaction data. The model can identify patterns and anomalies indicative of fraud, enabling proactive measures to protect customers. The secure deployment and role-based access control provided by MOMENTUM ensure sensitive financial data remains protected.</p><h3>Resume Matching</h3><p>Recruitment companies can utilize MOMENTUM to fine-tune an LLM for resume matching. The model can evaluate resumes against job descriptions, scoring them based on relevance and fit. This automated matching process helps recruiters quickly identify the best candidates, improving the hiring process and reducing time-to-hire.</p><h3>Contract Writing</h3><p>Legal firms can use MOMENTUM to develop an LLM fine-tuned for contract writing. The model can assist lawyers in drafting contracts by suggesting clauses, checking for consistency, and ensuring compliance with legal standards. This tool enhances productivity and accuracy, allowing legal professionals to focus on more complex tasks.</p><h3>Sentiment Analysis</h3><p>E-commerce companies can employ MOMENTUM to fine-tune an LLM for sentiment analysis on customer reviews and social media feedback. The model can categorize feedback into positive, negative, and neutral sentiments, providing insights into customer perceptions. This analysis helps companies make informed decisions on product improvements and marketing strategies.</p><h3>Translation Services</h3><p>Global enterprises can leverage MOMENTUM to fine-tune an LLM for translation services. The model can provide accurate translations across multiple languages, supporting international business operations. Continuous monitoring and updates ensure the translations remain relevant and contextually appropriate.</p><h3>Conclusion</h3><p>Accure’s MOMENTUM platform is a comprehensive solution for LLMOps, offering robust tools for data preparation, fine-tuning, testing, deployment, and interactive use. By streamlining and automating key aspects of the LLM lifecycle, MOMENTUM enables organizations to leverage the full potential of large language models, driving innovation and delivering value across various applications. To learn more about MOMENTUM visit <a href="https://accure.ai/">https://accure.ai</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c6364aac7ccc" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Handwritten Signature Extraction and Matching]]></title>
            <link>https://ansarisam.medium.com/handwritten-signature-extraction-and-matching-b85a27bb8b57?source=rss-9663a477bb27------2</link>
            <guid isPermaLink="false">https://medium.com/p/b85a27bb8b57</guid>
            <category><![CDATA[signature-matching]]></category>
            <category><![CDATA[fraud-prevention]]></category>
            <category><![CDATA[computer-vision]]></category>
            <category><![CDATA[compliance]]></category>
            <dc:creator><![CDATA[Sam Ansari]]></dc:creator>
            <pubDate>Mon, 20 Feb 2023 06:25:14 GMT</pubDate>
            <atom:updated>2025-08-01T18:05:02.635Z</atom:updated>
            <content:encoded><![CDATA[<p>Signature extraction and matching is a vital tool in preventing fraud, verifying identity, ensuring legal compliance, and improving efficiency in various industries. For example:</p><ol><li>Financial Services: Handwritten signature extraction and matching can be used by financial institutions to verify signatures on checks, loan documents, and other financial transactions. This helps prevent fraud and ensures that the signature belongs to the authorized person.</li><li>Government: Government agencies use signature extraction and matching to verify the authenticity of signatures on legal documents, passports, and other government-issued identification documents.</li><li>Healthcare: In the healthcare industry, handwritten signature extraction and matching can be used to verify patient signatures on consent forms and other medical documents.</li><li>Legal: In the legal industry, signature extraction and matching can be used to verify signatures on legal documents, such as contracts and agreements.</li><li>Education: Signature extraction and matching can be used by educational institutions to verify the identity of students on examination answer sheets, registration forms, and other documents.</li></ol><p>Signature extraction is a process of detecting handwritten signatures, locating its positions, and cropping the signature portions from a document. The location coordinates are used to crop the sections of the document to extract signatures. The following figures (1, 2 and 3) illustrated the signature extractions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/579/1*Tnra2WrDI01dlBXwBhSY3g.png" /><figcaption>Figure 1: Signatures extracted from the 1040 Tax form</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*acxEQkf56CXEiA8rwfcR4w.png" /><figcaption>Figure 2: Signature extracted from the Virginia Driver’s License</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LhNjVx_OPgrOhPBUS9dMig.png" /><figcaption>Figure 3: Signature extracted from the Indian PAN card</figcaption></figure><p>Signature matching is a process of comparing two or more signatures and obtaining a their match scores. Figure 4 shows the signature matching process.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hmCaykfLaTRyIHGqTI7RfA.png" /><figcaption>Figure 4: Signature matching with score. Keypoints are used to align the orientations of the two images.</figcaption></figure><p>Typically, the signature matching is a two step process:</p><p>Step 1: Signatures are first extracted from two or more scanned documents. The output from this extraction is typically cropped portions of signatures only.</p><p>Step 2: The cropped signatures are compared and match score is obtained.</p><p>However, in the case when signatures are already extracted by other means and cropped images are available, step 1 may be skipped and only step2 is needed to perform the comparison and matching.</p><p>In this article, we explore how to extract signatures from scanned documents and then perform matching using Restful APIs. To use the APIs, you need to have an account and security token. Here are the steps to do so:</p><ol><li>Sign up for an account at <a href="https://automate.accure.ai/register/">https://automate.accure.ai/register/</a> (if this is not working, please visit <a href="https://accure.ai/">https://accure.ai</a> and request access)</li><li>Login to <a href="https://automate.accure.ai/">https://automate.accure.ai/</a> using your username and password.</li><li>From the main menu panel (left side panel), click “APIs”. You may have to scroll down to see the link to “APIs”.</li><li>Key in your account password (the same password you use to login to your account), and click (do not hit enter key) “Reveal Key”.</li><li>API key will show up. Copy the key. You will need this key to perform the signature extraction and matching. Figure 5 shows a screenshot of the process of obtaining the API key.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*r3xCiVMc-y8MZSpytvTU7w.png" /><figcaption>Figure 5: Screenshot showing how to obtain security token or API key</figcaption></figure><h3>API Specifications</h3><p>Service Endpoint: <a href="https://automate.accure.ai/">https://automate.accure.ai/</a></p><h4>Signature Extraction</h4><p>Service Name: /v1/sign/extract</p><p>Curl Example:</p><pre>curl https://automate.accure.ai/v1/sign/extract -H &#39;Authorization: Token 123456765432123456&#39; -F &quot;image=@/home/user1/Desktop/test1.png&quot;</pre><p>In the above curl command, “image” is the path to the scanned document. Replace the token with the security token you obtained above (Figure 5).</p><p>Response Example:</p><pre>{<br>  &#39;image&#39;:&#39;1040EZForm.png&#39;,<br>  &#39;detections&#39;: [{<br>  &#39;x1&#39;: 2303.0,<br>  &#39;y1&#39;: 457.0,<br>  &#39;x2&#39;: 2671.0,<br>  &#39;y2&#39;: 536.0,<br>  &#39;confidence&#39;: 0.813737333,<br>  &#39;class_id&#39;: 1.0<br>}],<br>  &#39;transaction_id&#39;: &#39;1676846795913847019&#39;<br>}</pre><p>The response contains a list of detections containing the coordinates of the rectangular regions enclosing either signatures or logos.</p><p>(x1, y1) represent the coordinates of the top left corner and (x2,y2) represent the coordinates of the bottom left of the rectangular region containing the signature within the document.</p><p>confidence: is the combined probability of detection and localization of the signature.</p><p>class_id: 0 means logo and 1 means signature.</p><p>transaction_id: a unique transaction id representing the unique API call.</p><h4>Signature Matching</h4><p>Service Name: /v1/sign/match</p><p>Curl Example:</p><pre>curl https://automate.accure.ai/v1/sign/match -H &#39;Authorization: Token 123456765432123456&#39;<br>-F &quot;image1=@/home/user1/Desktop/test1.jpg&quot; -F &quot;image2=@/home/user1/Desktop/test2.jpg&quot;</pre><p>Response Example</p><pre>{<br>  &#39;transaction_id&#39;:&#39;8989891&#39;,<br>  &#39;image1&#39;:&#39;test.jpg&#39;,<br>  &#39;image2&#39;:&#39;test2.jpg&#39;,<br>  &#39;match_score&#39;:0.8,<br>  &#39;status_code&#39;: 200<br>}</pre><p>The response contains a match_score that ranges between 0 and 1. The score 1 means a perfect match and a 0 means no match.</p><p><strong>What is considered a good match?</strong></p><p>Depending on the level of noise in the signature image, the match_score varies, and it is up to the consumer of the API to decide where to put the threshold for a good match.</p><p>In the most common scenario, a match_score &gt; 0.7 is considered a good match. The higher the score, the better the match is.</p><p><strong>Status Codes</strong></p><p>200: Ok</p><p>401: Authentication failed</p><p>415: Invalid file format</p><p>405: Method Not Allowed</p><p>500: Failed. Request was not processed</p><h3>Conclusion</h3><p>The two APIs discussed above provide easy access to pre-trained computer vision based neural networks to perform signature extraction and matching. These APIs are available for free for any testing and evaluation. To learn more about the signature extraction and matching APIs, their usage, and more, visit this website <a href="https://accure.ai/signature-matching/">https://accure.ai/signature-matching</a>.</p><p>Although the neural network models are trained and regularly updated on a wide variety of documents, it is likely that the signature extraction will fail in document types the model has never seen before. Also, the scanning quality, image clarity, and several other factors influence the accuracy of such extraction.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b85a27bb8b57" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>